0.1.4

Supertonic 3 TTS App v0.1.4

On-device neural text-to-speech for Windows. No cloud, no API keys, no data leaves your machine.

Downloads

File	Description
supertonic-3-tts-app_v0.1.4_x64-setup.exe	Portable NSIS installer - no UAC, no registry, no admin required
supertonic-3-tts-app_v0.1.4_x64-portable.exe	Standalone portable executable (~7.3 MB, UPX compressed)

Requirements

• Windows 10/11 (x64)
• WebView2 (pre-installed on Windows 11; Windows 10 may prompt to install)
• Internet connection on first run (to download ONNX models from HuggingFace - cached locally after)

What's New in v0.1.4

• Multi-threaded CPU inference support using ONNX Runtime Web, WASM threads, and SharedArrayBuffer
• CPU mode now auto-detects logical CPU threads from the native Tauri backend
• CPU usage slider added in Settings - choose 10% to 100% CPU usage, default 75%
• App volume slider added in Settings - default 80%, applies to generated audio, history playback, and voice previews
• GPU and CPU production loading fixed by bundling ONNX Runtime Web .mjs and .wasm assets at stable paths
• Generated audio panel redesigned with clearer playback, waveform, stats, WAV save, MP3 save, and bitrate controls
• Saved WAV and MP3 filenames now include date/time to prevent overwriting previous exports
• Audio history behavior fixed - current generation stays in the main view, previous generations move into history
• History now shows up to 6 previous generations
• Playback bug fixed where the main player could play the previous generation after unloading and generating new text
• "Normalize audio" renamed to "Normalize output"
• CPU UI freeze warning removed from all languages
• Version updated to v0.1.4 in app footer, settings, package metadata, and installer metadata
• English and Chinese READMEs updated with the new CPU threading, usage slider, volume, export, and history behavior

Features

• 10 built-in English voices - 5 female (F1-F5) and 5 male (M1-M5)
• Custom voice cloning - load .json voice style embeddings trained with supertonic_embeddings_trainer
• GPU acceleration - WebGPU with automatic CPU fallback
• Multi-threaded CPU fallback - WASM SIMD + SharedArrayBuffer with adjustable CPU usage
• File input - .txt, .pdf, .docx, .md (up to 20 MB)
• Steps control - 5 (speed) to 16 (quality)
• Speed control - 0.5x to 2.0x
• Normalize output - peak normalization toggle (playback, history, downloads)
• Waveform seekbar - click to seek during playback
• Audio history - current generation plus up to 6 previous generations with metadata
• Export - timestamped WAV or MP3 (128 / 192 / 320 kbps)
• App volume control - 0% to 100%
• Dark / Light theme with 5 accent colors
• Resizable two-column layout with UI scaling (80%-150%)
• 7 UI languages - English, 中文, Español, Français, Ελληνικά, Русский, 日本語
• SHA-256 verified model downloads

Powered by

• Supertone Supertonic 3 - neural TTS engine
• HuggingFace Supertone/supertonic-3 - ONNX models (OpenRAIL-M License)
• Tauri 2 + React 19 + ONNX Runtime Web

0.1.3

Supertonic 3 TTS App v0.1.3

On-device neural text-to-speech for Windows. No cloud, no API keys, no data leaves your machine.

Downloads

File	Description
supertonic-3-tts-app_v0.1.3_x64-setup.exe	Portable NSIS installer — no UAC, no registry, no admin required
supertonic-3-tts-app_v0.1.3_x64-portable.exe	Standalone portable executable (~7.2 MB, UPX compressed)

Requirements

• Windows 10/11 (x64)
• WebView2 (pre-installed on Windows 11; Windows 10 may prompt to install)
• Internet connection on first run (to download ONNX models from HuggingFace — cached locally after)

What's New in v0.1.3

• 7 UI languages — English, 中文, Español, Français, Ελληνικά, Русский, 日本語 (selectable in Settings)
• Normalize audio now applies to history playback, WAV download, and MP3 download (not just the output player)
• Voice clone tooltip redesigned — shorter link text, stays within panel bounds
• App renamed to Supertonic 3 TTS App across all UI, installer, window title, and repo branding
• GitHub links updated to Saganaki22/Supertonic_3_TTS_App
• Custom voice ? link now points to supertonic_embeddings_trainer
• WAV download now applies normalize audio setting
• Language dropdown in settings matches app theme (dark/light)
• Version displayed correctly as v0.1.3 in footer and settings

Features

• 10 built-in English voices — 5 female (F1–F5) and 5 male (M1–M5)
• Custom voice cloning — load .json voice style embeddings trained with supertonic_embeddings_trainer
• GPU acceleration — WebGPU with automatic CPU (WASM SIMD) fallback
• File input — .txt, .pdf, .docx, .md (up to 20 MB)
• Steps control — 5 (speed) to 16 (quality)
• Speed control — 0.5× to 2.0×
• Audio normalization — peak normalization toggle (playback, history, downloads)
• Waveform seekbar — click to seek during playback
• Audio history — last 10 generations with metadata
• Export — WAV or MP3 (128 / 192 / 320 kbps)
• Dark / Light theme with 5 accent colors
• Resizable two-column layout with UI scaling (80%–150%)
• SHA-256 verified model downloads

Powered by

• Supertone Supertonic 3 — neural TTS engine
• HuggingFace Supertone/supertonic-3 — ONNX models (OpenRAIL-M License)
• Tauri 2 + React 19 + ONNX Runtime Web

0.1.2

Supertonic 3 TTS App v0.1.2

On-device neural text-to-speech for Windows. No cloud, no API keys, no data leaves your machine.

Downloads

File	Description
Supertonic.3.TTS.App_v0.1.2_x64-setup.exe	NSIS installer — no UAC, no registry, no admin required
supertonic-3-tts-app_v0.1.2_x64-portable.exe	Standalone portable executable (~7.2 MB, UPX compressed)

Requirements

• Windows 10/11 (x64)
• WebView2 (pre-installed on Windows 11; Windows 10 may prompt to install)
• Internet connection on first run (to download ONNX models from HuggingFace — cached locally after)

Features

• 10 built-in English voices — 5 female (F1–F5) and 5 male (M1–M5)
• Custom voice cloning — load .json voice style embeddings trained with supertonic_embeddings_trainer
• GPU acceleration — WebGPU with automatic CPU (WASM SIMD) fallback
• File input — .txt, .pdf, .docx, .md (up to 20 MB)
• Steps control — 5 (speed) to 16 (quality)
• Speed control — 0.5× to 2.0×
• Audio normalization — peak normalization toggle (playback, history, downloads)
• Waveform seekbar — click to seek during playback
• Audio history — last 10 generations with metadata
• Export — WAV or MP3 (128 / 192 / 320 kbps)
• Dark / Light theme with 5 accent colors
• Resizable two-column layout with UI scaling (80%–150%)
• SHA-256 verified model downloads

Powered by

• Supertone Supertonic 3 — neural TTS engine
• HuggingFace Supertone/supertonic-3 — ONNX models (OpenRAIL-M License)
• Tauri 2 + React 19 + ONNX Runtime Web